Research On Target Tracking Algorithm Based On Compressed Sensing On The Edge

With the construction of smart cities and the development of UAV technology,using UAV video monitoring system to track targets has become a hotspot on the researches.But it needs a lot of computing resources to process the video of multiple cameras at the same time,and the computing resources of the network edge of video monitoring system are very limited.Considering the reasons of limited bandwidth and protecting user privacy,it is not practical to upload this massive data to the cloud data center for processing.In order to effectively use the limited computing resources on the edge of the network to track the target,the work of this paper is as follows:In this paper,a new target tracking algorithm is proposed,which based on compressed sensing on the edge of UAV video monitoring system network.The algorithm includes three parts: virtual grid algorithm,target detection algorithm based on compressed sensing in space domain and target detection algorithm based on compressed sensing in time domain.By dividing the virtual grid to transform the dynamic video monitoring problem into a static problem and sample the video data in the virtual grid.By processing the sampled data with packet sampling and compressed sensing to recover the complete target information.The target appearance information is abstractly calculated as the probability that the target appears in the monitoring system of each time unit,the time occurrence probability value is sampled,and the target appearance time range in the time domain is recovered using compressed sensing according to the sampling result.At this time,the position information of the target in the spatial domain and the time range of the appearance in the time domain are obtained,and the target tracking is completed.The focus of this algorithm is not on target detection.Finally,a great quantity of simulation experiments are designed to evaluate the recovery accuracy of the algorithm.The experimental results show that the algorithm can recover the target detection information with high accuracy,and can only process less than ten percent of the video frame data to get the complete target track information,so as to effectively use the edge computing resources to achieve efficient target tracking.